handle checking for file better

README.md

@@ -40,14 +40,18 @@ $EDITOR Crawler.toml
 - [x] Allow for storing asynchronously - dropping the "links to" logic fixes this need
 - [x] Control crawler via config file (no recompliation needed)
 
-3/17/25: Took >1hr to crawl 100 pages
+### Feats
 
-3/19/25: Took 20min to crawl 1000 pages
+3/17/25: Took >1hr to crawl 100 pages.
+
+3/19/25: Took 20min to crawl 1000 pages.
 This ment we stored 1000 pages, 142,997 urls, and 1,425,798 links between the two.
 
-3/20/25: Took 5min to crawl 1000 pages
+3/20/25: Took 5min to crawl 1000 pages.
 
-3/21/25: Took 3min to crawl 1000 pages
+3/21/25: Took 3min to crawl 1000 pages.
+
+7/.../25: Downloaded just shy of 12TB of data from a remote server.
 
 # About
 

src/db.rs

@@ -59,7 +59,6 @@ impl Website {
         counter!(STORE).increment(1);
         let mut things = Vec::with_capacity(all.len());
 
-        // FIXME fails *sometimes* because "Resource Busy"
         match db
             .query(
                 "INSERT INTO website $array
@@ -103,7 +102,7 @@ pub async fn get_next(db: &Surreal<Client>, config: &Config) -> Option<Website>
             Ok(ok) => ok,
             Err(_err) => {
                 // basically just CSMA/CA
-                let delay = rand::random_range(10..500);
+                let delay = rand::random_range(10..10_000);
                 sleep(Duration::from_millis(delay)).await;
                 fails += 1;
                 // Don't get stuck here forever, failing...
@@ -147,9 +146,9 @@ pub async fn connect(config: &Config) -> surrealdb::Result<Surreal<Client>> {
         .await?;
 
     let setup = include_bytes!("setup.surql");
-    let file = setup.iter().map(|c| *c as char).collect::<String>();
+    let init_commands = setup.iter().map(|c| *c as char).collect::<String>();
 
-    db.query(file)
+    db.query(init_commands)
         .await
         .expect("Failed to setup surreal tables.");
 

src/filesystem.rs

@@ -34,14 +34,46 @@ pub fn as_path(url: &Url, content_type: &HeaderValue) -> PathBuf {
     url_path
 }
 
+pub async fn check_file_length(file: &PathBuf) -> Option<u64> {
+    match tokio::fs::OpenOptions::new()
+        .write(false)
+        .read(true)
+        .create(false)
+        .open(file).await
+    {
+        Ok(file) => {
+            match file.metadata().await {
+                Ok(meta) => {
+                    return Some(meta.len())
+                },
+                Err(err) => {
+                    error!("Failed to get metadata. {}", err)
+                },
+            }
+        },
+        Err(err) => {
+            match err.kind() {
+                ErrorKind::NotFound => {/* ignore */},
+                _ => warn!("Failed to open file to check length. {:?} {}", file, err),
+            }
+        },
+    }
+    None
+
+}
+
 pub async fn init(filename: &PathBuf) -> Option<fs::File> {
     let file = async || tokio::fs::OpenOptions::new()
-        .append(true)
+        .write(true)
+        .append(false)
         .create(true)
         .open(&filename).await;
 
     match file().await {
-        Ok(ok) => Some(ok),
+        Ok(ok) => {
+            trace!("Initialized file {}", filename.to_str().unwrap_or("N/A"));
+            Some(ok)
+        },
         Err(err) => {
             // the file/folder isn't found
             if err.kind() == ErrorKind::NotFound {
@@ -55,6 +87,14 @@ pub async fn init(filename: &PathBuf) -> Option<fs::File> {
                 } else {
                     error!("Couldn't get file's parents: {:?}", &filename);
                 }
+            } else if err.kind() == ErrorKind::NotADirectory {
+                // Example:
+                // 1. example.com/user
+                // 2. example.com/user/post
+                // If file 1 exists it will prevent file 2 from existing
+                // FIXME
+
+                error!("One of the parent directories is actually a file...")
             } else {
                 error!("File open error: {err} {:?}", filename);
             }

src/main.rs

@@ -5,16 +5,27 @@
 extern crate html5ever;
 
 use std::{
-    collections::HashSet, fs::File, io::Read, sync::{Arc, LazyLock}
+    collections::HashSet,
+    fs::File,
+    io::Read,
+    sync::{Arc, LazyLock},
 };
 
-use futures_util::StreamExt;
-use opentelemetry::{global::{self}, metrics::{Counter, Meter, UpDownCounter}};
-use opentelemetry_otlp::{Protocol, WithExportConfig};
 use db::{connect, Website};
+use futures_util::StreamExt;
+use opentelemetry::{
+    global::{self},
+    metrics::{Counter, Meter, UpDownCounter},
+};
+use opentelemetry_otlp::{Protocol, WithExportConfig};
+use opentelemetry_sdk::{metrics::SdkMeterProvider, trace::SdkTracerProvider};
 use serde::Deserialize;
 use surrealdb::{engine::remote::ws::Client, Surreal};
-use tokio::{io::{AsyncWriteExt, BufWriter}, sync::RwLock, task::JoinSet};
+use tokio::{
+    io::{AsyncReadExt, AsyncWriteExt, BufWriter},
+    sync::RwLock,
+    task::JoinSet,
+};
 use tracing::{debug, error, info, instrument, level_filters::LevelFilter, trace, warn};
 use tracing_subscriber::{fmt, layer::SubscriberExt, EnvFilter, Layer, Registry};
 
@@ -25,36 +36,29 @@ mod filesystem;
 mod parser;
 
 static METER: LazyLock<Meter> = LazyLock::new(|| global::meter("Internet_Mapper"));
-static BEING_PROCESSED: LazyLock<UpDownCounter<i64>> = LazyLock::new(||
+static BATCH_SIZE: LazyLock<Counter<u64>> =
+    LazyLock::new(|| METER.u64_counter("crawler_batch_size").build());
+static BEING_PROCESSED: LazyLock<UpDownCounter<i64>> = LazyLock::new(|| {
     METER
         .i64_up_down_counter("crawler_pages_being_processed")
         .build()
-);
+});
-static BEING_PARSED: LazyLock<UpDownCounter<i64>> = LazyLock::new(||
+static BEING_PARSED: LazyLock<UpDownCounter<i64>> = LazyLock::new(|| {
     METER
         .i64_up_down_counter("crawler_pages_being_parsed")
         .build()
-);
+});
-static BEING_STREAMED: LazyLock<UpDownCounter<i64>> = LazyLock::new(|| 
+static BEING_STREAMED: LazyLock<UpDownCounter<i64>> = LazyLock::new(|| {
     METER
         .i64_up_down_counter("crawler_pages_being_streamed")
         .build()
-);
+});
-static GET_IN_FLIGHT: LazyLock<UpDownCounter<i64>> = LazyLock::new(|| 
+static GET_IN_FLIGHT: LazyLock<UpDownCounter<i64>> =
-    METER
+    LazyLock::new(|| METER.i64_up_down_counter("crawler_gets_in_flight").build());
-        .i64_up_down_counter("crawler_gets_in_flight")
+static TOTAL_BYTES_DOWN: LazyLock<Counter<u64>> =
-        .build()
+    LazyLock::new(|| METER.u64_counter("crawler_total_bytes_down").build());
-);
+static SITES_CRAWLED: LazyLock<Counter<u64>> =
-static TOTAL_BYTES_DOWN: LazyLock<Counter<u64>> = LazyLock::new(|| 
+    LazyLock::new(|| METER.u64_counter("crawler_total_sites_crawled").build());
-    METER
-        .u64_counter("crawler_total_bytes_down")
-        .build()
-);
-static SITES_CRAWLED: LazyLock<Counter<u64>> = LazyLock::new(|| 
-    METER
-        .u64_counter("crawler_total_sites_crawled")
-        .build()
-);
 
 static CONFIG: LazyLock<Config> = LazyLock::new(|| {
     let mut file = File::open("./Crawler.toml").expect("Failed to read Crawler.toml");
@@ -91,9 +95,11 @@ async fn main() {
     println!("Logs and metrics are provided to the Grafana dashboard");
 
     // Start TRACE / LOGGING / METRICS
-    load_tracing(&CONFIG);
+    load_logging(&CONFIG); // this seems to be working ok
-    load_logging(&CONFIG);
+    global::set_tracer_provider(load_tracing(&CONFIG));
-    load_metrics(&CONFIG);
+    global::set_meter_provider(load_metrics(&CONFIG));
+
+    BATCH_SIZE.add(CONFIG.batch_size as u64, &[]);
 
     // When getting uncrawled pages, name must contain this variable. "" will effectively get ignored.
     // let crawl_filter = "en.wikipedia.org/";
@@ -122,15 +128,40 @@ async fn main() {
     // let mut main_loop_span= TRACER.start("Main-Loop");
     let mut futures = JoinSet::new();
     for _ in 0..CONFIG.batch_size {
-        futures.spawn(process_single_thread(&CONFIG, db.clone(), reqwest.clone(), crawled.clone()));
+        futures.spawn(process_single_thread(
+            &CONFIG,
+            db.clone(),
+            reqwest.clone(),
+            crawled.clone(),
+        ));
     }
+
+    while let Some(_) = futures.join_next().await {
+        // Budget - Threads - This thread (1)
+        // Would roughly be the acceptable amount at which a thread should exit
+        if *(crawled.read().await) < CONFIG.budget - CONFIG.batch_size - 1 {
+            warn!("Thread terminated early, restarting");
+            futures.spawn(process_single_thread(
+                &CONFIG,
+                db.clone(),
+                reqwest.clone(),
+                crawled.clone(),
+            ));
+        }
+    }
+
     futures.join_all().await;
     // main_loop_span.end();
 
     info!("Done");
 }
 
-async fn process_single_thread(config: &Config, db: Surreal<Client>, reqwest: reqwest::Client, crawled: Arc<RwLock<usize>>) {
+async fn process_single_thread(
+    config: &Config,
+    db: Surreal<Client>,
+    reqwest: reqwest::Client,
+    crawled: Arc<RwLock<usize>>,
+) {
     while *(crawled.read().await) < config.budget {
         let uncrawled = get_next(&db.clone(), &config).await;
         match uncrawled {
@@ -140,11 +171,11 @@ async fn process_single_thread(config: &Config, db: Surreal<Client>, reqwest: re
                 // Somehow this write doesn't hang on the while's read?
                 let mut c = crawled.write().await;
                 *c += 1;
-            },
+            }
             None => {
                 warn!("fn::get_next() returned None");
                 return;
-            },
+            }
         }
     }
 }
@@ -153,7 +184,6 @@ async fn process_single_thread(config: &Config, db: Surreal<Client>, reqwest: re
 /// Downloads and crawls and stores a webpage.
 /// It is acceptable to clone `db`, `reqwest`, and `s3` because they all use `Arc`s internally. - Noted by Oliver
 async fn process(mut site: Website, db: Surreal<Client>, reqwest: reqwest::Client) {
-
     // METRICS
     debug!(url = &site.site.as_str(), "Process: {}", &site.site);
     BEING_PROCESSED.add(1, &[]);
@@ -165,6 +195,8 @@ async fn process(mut site: Website, db: Surreal<Client>, reqwest: reqwest::Clien
     // Send the http request (get)
     GET_IN_FLIGHT.add(1, &[]);
     if let Ok(response) = request_builder.send().await {
+        let mut skip_download = false;
+
         GET_IN_FLIGHT.add(-1, &[]);
 
         let headers = response.headers();
@@ -177,69 +209,113 @@ async fn process(mut site: Website, db: Surreal<Client>, reqwest: reqwest::Clien
         let CT = headers.get("Content-Type");
         let ct = headers.get("content-type");
 
-        let ct = match (CT,ct) {
+        let ct = match (CT, ct) {
             (None, None) => {
-                warn!("Server did not respond with Content-Type header. Url: {} Headers: ({:?})", site.site.to_string(), headers);
+                warn!(
-                return
+                    "Server did not respond with Content-Type header. Url: {} Headers: ({:?})",
-            },
+                    site.site.to_string(),
+                    headers
+                );
+                return;
+            }
             (None, Some(a)) => a,
             (Some(a), None) => a,
             (Some(a), Some(_)) => a,
         };
 
         // create filepath (handles / -> /index.html)
-        let path = filesystem::as_path(&site.site, ct);
+        let real_path = filesystem::as_path(&site.site, ct);
-        let mut tmp_path= path.clone();
+        let mut tmp_path = real_path.clone();
         if !(tmp_path.add_extension("crawl_temp")) {
             warn!("Failed to add extension to file");
             // fallback ig
             tmp_path = tmp_path.with_extension("crawl_temp");
         }
 
+        // CODE FOR UPDATING DOWNLOADED CONTENT:
+        // Check the Content-Length header (we assume the server is telling the truth) (I don't see
+        // a reason for it to lie in this case).
+        // And see if the file on the disk is the same length.
+        // Yes, technically this isn't perfect, but the other option is storing ETags, which I
+        // don't want to do right now.
+        if let Some(len) = headers.get("Content-Length") {
+            if let Ok(s) = len.to_str() {
+                // length is in bytes
+                if let Ok(len) = s.parse::<u64>() {
+                    if let Some(disk_len) = filesystem::check_file_length(&real_path).await {
+                        if disk_len == len {
+                            skip_download = true;
+                        }
+                    } else {
+                        // File not found (or other error).
+                        // Program will continue on it's way, downloading content.
+                    }
+                }
+            }
+        }
+
         // make sure that the file is good to go
         if let Some(file) = filesystem::init(&tmp_path).await {
             // Get body from response
             // stream the response onto the disk
             let mut stream = response.bytes_stream();
 
-            let should_parse = path.to_string_lossy().ends_with(".html");
+            let should_parse = real_path.to_string_lossy().ends_with(".html");
-            let mut writer = BufWriter::new(file);
+
             let mut buf: Vec<u8> = Vec::new();
 
-            // Write file to disk
+            if skip_download && should_parse {
-            trace!("Writing at: {:?}", tmp_path);
+                // since we are skipping the download we will just read the file off the disk to
-            BEING_STREAMED.add(1, &[]);
+                // parse it
-            // let mut stream_span = TRACER.start("Stream");
+                if let Ok(mut file) = tokio::fs::OpenOptions::new() 
-            while let Some(data) = stream.next().await {
+                    .read(true)
-                match data {
+                    .open(&real_path).await
-                    Ok(data) => {
+                {
-                        TOTAL_BYTES_DOWN.add(data.len() as u64, &[]);
+                    if let Err(err) = file.read_to_end(&mut buf).await {
-                        let _ = writer.write_all(&data).await;
+                        warn!("Failed to read file off disk for parsing, {}", err);
-                        // If we are going to parse this file later, we will save it
+                    }
-                        // into memory as well as the disk.
-                        // We do this because the data here might be incomplete
-                        if should_parse {
-                            data.iter().for_each(|f| buf.push(*f));
-                        }
-                    },
-                    Err(err) => {
-                        error!("{err}")
-                    },
                 }
             } 
-            let _ = writer.flush().await;
-            // rename the temp file into the real file name
-            if let Err(err) = tokio::fs::rename(&tmp_path, &path).await {
-                error!(
-                    from = &*tmp_path.to_string_lossy(),
-                    to = &*path.to_string_lossy(),
-                    "Error renaming file: {}",
-                    err
-                );
-            }
 
-            // stream_span.end();
+            // !!!DOWNLOADING TIME!!!
-            BEING_STREAMED.add(-1, &[]);
+            if !skip_download {
+                let mut writer = BufWriter::new(file);
+
+                // Write file to disk
+                trace!("Writing at: {:?}", tmp_path);
+                BEING_STREAMED.add(1, &[]);
+                // let mut stream_span = TRACER.start("Stream");
+                while let Some(data) = stream.next().await {
+                    match data {
+                        Ok(data) => {
+                            TOTAL_BYTES_DOWN.add(data.len() as u64, &[]);
+                            let _ = writer.write_all(&data).await;
+                            // If we are going to parse this file later, we will save it
+                            // into memory as well as the disk.
+                            // We do this because the data here might be incomplete
+                            if should_parse {
+                                data.iter().for_each(|f| buf.push(*f));
+                            }
+                        }
+                        Err(err) => {
+                            error!("{err}")
+                        }
+                    }
+                }
+                let _ = writer.flush().await;
+                // rename the temp file into the real file name
+                if let Err(err) = tokio::fs::rename(&tmp_path, &real_path).await {
+                    error!(
+                        from = &*tmp_path.to_string_lossy(),
+                        to = &*real_path.to_string_lossy(),
+                        "Error renaming file: {}",
+                        err
+                    );
+                }
+
+                // stream_span.end();
+                BEING_STREAMED.add(-1, &[]);
+            }
 
             // (If needed) Parse the file
             if should_parse {
@@ -263,25 +339,23 @@ async fn process(mut site: Website, db: Surreal<Client>, reqwest: reqwest::Clien
                 // parsing_span.end();
                 BEING_PARSED.add(-1, &[]);
             } else {
-                trace!("Did not parse: {}", site.site.as_str());
+                trace!(url = site.site.as_str(), "Parse = False");
             }
 
             // update self in db
             site.crawled = true;
             site.status_code = code.as_u16();
             Website::store_all(vec![site.clone()], &db).await;
-        } else {
-            error!("File failed to cooperate: {:?}", path);
         }
     } else {
-        error!("Failed to get: {}", &site.site);
+        error!(url = site.site.as_str(), "Failed to get: {}", &site.site);
     }
 
     // process_span.end();
     BEING_PROCESSED.add(-1, &[]);
 }
 
-fn load_tracing(config: &Config) {
+fn load_tracing(config: &Config) -> SdkTracerProvider {
     // Send spans to Alloy (which will send them to Tempo)
     let otlp_span = opentelemetry_otlp::SpanExporter::builder()
         .with_tonic()
@@ -291,16 +365,15 @@ fn load_tracing(config: &Config) {
     let tracer_provider = opentelemetry_sdk::trace::SdkTracerProvider::builder()
         .with_simple_exporter(otlp_span)
         .build();
-
+    tracer_provider
-    global::set_tracer_provider(tracer_provider);
 }
 
 fn load_logging(config: &Config) {
-//    let otlp_log = opentelemetry_otlp::LogExporter::builder()
+    //    let otlp_log = opentelemetry_otlp::LogExporter::builder()
-//        .with_tonic()
+    //        .with_tonic()
-//        .with_endpoint(endpoint)
+    //        .with_endpoint(endpoint)
-//        .build()
+    //        .build()
-//        .unwrap();
+    //        .unwrap();
     // let logger_provider = opentelemetry_sdk::logs::SdkLoggerProvider::builder()
     //    .with_simple_exporter(otlp_log)
     //    .build();
@@ -321,13 +394,13 @@ fn load_logging(config: &Config) {
             .with_file(true)
             .json()
             .with_writer(writer)
-            .with_filter(filter)
+            .with_filter(filter),
     );
 
     tracing::subscriber::set_global_default(registry).expect("Failed to set default subscriber");
 }
 
-fn load_metrics(config: &Config) {
+fn load_metrics(config: &Config) -> SdkMeterProvider {
     //  Send metrics to Prometheus
     let otlp_metrics = opentelemetry_otlp::MetricExporter::builder()
         .with_http()
@@ -338,8 +411,5 @@ fn load_metrics(config: &Config) {
     let metrics_provider = opentelemetry_sdk::metrics::SdkMeterProvider::builder()
         .with_periodic_exporter(otlp_metrics) // default delay is 60s, turn down to like 15
         .build();
-
+    metrics_provider
-    global::set_meter_provider(metrics_provider);
 }
-
-

src/setup.surql

@@ -14,3 +14,5 @@ DEFINE FUNCTION OVERWRITE fn::get_next($filter: string) {
     UPDATE $site.id SET processing = true;
     RETURN $site
 };
+
+UPDATE website SET processing = false WHERE processing = true;